Research ArticleStatisticalEvaluationof theMaterial-SourceEffect on theDuctilityand Elastic Recovery (ER) of Plant-Mix Extracted Asphalt-Binders

Lubinda F. Walubita,1 Gilberto Martinez-Arguelles ,2 Harshavardhan R. Chunduri,1

Jose G. Gonzalez Hernandez,2 and Luis Fuentes 2

1TI-�e Texas A&M University System, College Station, Texas, USA2Department of Civil & Environmental Engineering, Universidad del Norte (UniNorte), Barranquilla, Colombia

Correspondence should be addressed to Gilberto Martinez-Arguelles; garguelles@uninorte.edu.co

Received 23 June 2020; Revised 22 September 2020; Accepted 8 October 2020; Published 28 October 2020

Academic Editor: John Kechagias

Copyright © 2020 Lubinda F. Walubita et al. ,is is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work isproperly cited.

,is study was conducted to quantitatively and statistically evaluate the effects of material source on the ductility of asphalt-binders, measured in terms of the elastic recovery (ER) property. ,e ER data used in the study were excerpted from the Texasflexible pavements and overlays database, namely, the Texas Data Storage System (DSS), covering plant-mix extracted PG XX-22asphalt-binders (i.e., rolling thin film oven (RTFO) residues) from 20 different sources and measured using the Ductilometer testat 10°C. ,e findings of the study indicated that material source has an impact on the ER property of asphalt-binders. Statisticallysignificant differences were observed among some sources and suppliers that reported the same low-temperature asphalt-bindertype/grade (i.e., PG XX-22). Overall, the study contributes to enriching the literature on the material-source effects on asphalt-binders’ ER properties, consistency, variability, and data quality. In particular, the study highlights the sensitivity nature of theasphalt-binder ER parameter to material-source effects.

1. Introduction

Among many other influencing factors, the qualityconsistency and properties of asphalt-binders are de-pendent on the production process and, subsequently, thecorresponding asphalt-binder sources and suppliers[1–3]. Consequently, different asphalt-binders fromdifferent sources and suppliers, even those classified withthe same type/grade, may thus exhibit different rheo-logical and viscoelastic properties that have an inherentimpact on the resultant hot-mix asphalt (HMA) prop-erties and the overall field performance [3–5]. ,is isfurther exacerbated by the current asphalt-binder pro-duction methods/processes that have changed signifi-cantly, among others, due to technical, economic, andenvironmental evolutions [6, 7]. ,us, studies orientedtowards enhancing the chemical, physical, rheological,and viscoelastic properties of asphalt-binders are para-mount [6–9] to optimize the HMA durability and

mitigate against premature pavement failures such ascracking [7].

Cracking is one of the major distresses that undesirablyreduce the durability and long-term performance of HMApavements [10, 11]. Asphalt-binder, including type/gradeand its volumetric content in the HMA mix, significantlyinfluences the cracking resistance properties of HMA mixesand ultimately the cracking performance in the field [12].,us, having good-quality asphalt-binders and adequatelycharacterizing their viscoelastic properties such as ductilityand elastic recovery (ER) that are related to cracking per-formance are imperative [13]. However, as mentionedabove, asphalt-binders classified with the same types/gradesbut obtained from different sources and suppliers couldexhibit different viscoelastic properties with different per-formance impacts on both the resultant HMA mixes andpavements in the field [3], hence the need to study thematerial-source effects on the asphalt-binders’ ductility andER properties.

HindawiAdvances in Civil EngineeringVolume 2020, Article ID 8851691, 12 pageshttps://doi.org/10.1155/2020/8851691

mailto:garguelles@uninorte.edu.cohttps://orcid.org/0000-0002-3419-0012https://orcid.org/0000-0002-7811-8821https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://doi.org/10.1155/2020/8851691

,e ductility of asphalt-binders, as measured based onthe ASTM D113-17 test standard [14], provides a goodindicator for the long-term durability performance of as-phalt-binders, allowing for the mitigation against cracking[15]. ,e literature reports that the ductility of asphalt-binders that recovered from HMA pavements correlateswith cracking failure [15]. Some field tests have also indi-cated that ductility measured at low temperatures is a goodindicator of the age-related cracking of asphalt-binders [16].Additionally, ductility is one of the traditionally used testmethods to characterize the asphalt-binder viscoelastic re-sponse and one of the primary requirements in the pene-tration grading specification of asphalt-binders [17].However, some of the literature reviewed have suggestedbeing cautious when selecting the ductility test loadingparameters to ensure that the resultant strain levels arerepresentative and simulative of the loading that typicallyoccurs in the HMA pavement during its service life in thefield [18, 19]. Similarly, some literature also advise cautionwhen analyzing and interpreting the ductility results as theycontend that ductility is an empirical property in terms of itsrelationship to the fundamental HMA material properties[20, 21]. Nonetheless, the ductility parameter continues to beused in many countries and is still used to provide aquantitative estimation and approximation of the asphalt-binder’s elastic properties, potential to recover after elon-gation or when subjected to tensile loading, and the resultantHMA’s cracking resistance potential [3, 20].

,e Superpave Performance Grading (PG) systemintroduced the ER test as one of the methods to assess theasphalt-binder property related to HMA cracking (fa-tigue) performance [22, 23]. ,is test evaluates andquantifies the elastic properties of asphalt-binders bymeasuring the amount of recoverable ability of the as-phalt-binder after elasticity deformation. ,e recoverypotential of the asphalt-binder is fundamentally seen as aself-recovery ability of the material, during which thedistress level decreases and the performance of the as-phalt-binder enhances with recovery time [19].

In general, as the asphalt-binder ages, it loses itsductility and self-recovery properties [19]. By and large,the recovery (ER) properties of asphalt-binders are con-sidered as the fundamental properties for correlation withHMA performance, particularly with respect to cracking[18]. However, the ER test is considered as a subjective testbecause it depends on an eyeball estimate and great care isneeded during handling, pouring, and trimming of theasphalt-binder specimen to ensure reliability of the testdata [24].

,e ER property of the asphalt-binder is inherentlyrelated to its chemical properties such as asphaltenes, resins,and oils that are source-dependent [5, 7, 8, 15, 25]. In thisstudy, a comprehensive statistical analysis was conducted toevaluate and quantify the effects of material source on the ERproperties of asphalt-binders. Analysis of Variance(ANOVA) and Tukey´s Honestly Significant Difference(Tukey´s HSD) statistical methods were used to compara-tively evaluate up to 20 different sources/suppliers of as-phalt-binders covering PG 64-22 and PG 76-22 asphalt-

binders, all extracted from plant-mix materials that weredirectly hauled from field construction sites [26].

In the subsequent sections of the paper, a review of theliterature is presented followed by the study matrix plan, testresults, statistical analysis, and synthesis of the findings. ,epaper finally concludes with a summary of key findings andhighlights of the research significance along with recom-mendations for future studies.

2. Literature Review

Many studies have been conducted on the variability anddifferences in the properties of asphalt-binders and therelation with HMA field performance has been studied bymany researchers [2, 4, 5, 27]. However, the literaturereviewed is limited with respect to studies on the impacts ofdifferent sources and suppliers on the ductility and ERproperties of asphalt-binders. Khan et al. [20] used a total of108 individual ductility tests (AASHTO T51-09) to measurethe ductility properties of 54 asphalt-binders hauled fromdifferent construction locations (i.e., sources) in Ontario,Canada. ,e test results yielded a range of ductility valuesbetween 14.3 and 161.3mm, which evidently represents alarge data variability. Reproducibility of the test resultsshowed a coefficient of variation (CoV) of 3.5% and a pooledstandard deviation (σp/range) value of 2.9mm for a singleoperator test. On the other hand, Alvarez et al. [3] alsomeasured the ductility property of 18 different Pen 60-70asphalt-binders, all from one refinery supplier in Colombia.,eir findings indicated no differences as all the asphalt-binders registered rupture at 148mm, which was themaximum distance limit of the testing equipment used; thus,the results could not provide any useful information on thevariability or reproducibility of the asphalt-binders tested.

Zhang et al. [10, 28] conducted some laboratory studiesto measure and characterize the ER properties of plant-mixextracted asphalt-binders relative to the HMA fractureproperties as a function of material source. In total, 11 Texasasphalt-binders from different sources and suppliers werecomparatively evaluated. ,e Ductilometer test results at10°C yielded an ER range of 23% to 49% for PG 64-22 and59% to 70% for PG 76-22 asphalt-binders, respectively,which clearly presents a huge difference and variabilityamong the different sources, for the asphalt-binders with thesame low-temperature grade of −22, i.e., PG XX-22.

2.1. Variability and Asphalt-Binder Source Effects. From theliterature reviewed above [1, 3, 10, 28], it can be inferred thatasphalt-binders have variability in terms of the rheologicaland viscoelastic (ER) properties that could be potentiallysource/supplier related. To further enrich the literature, thisstudy employed statistical methods to evaluate and quantifythe material-source effect on the asphalt-binders’ ERproperty. In particular, the study focused on plant-mixextracted asphalt-binder and the ER parameter measuredusing the Ductilometer device and covered two commonlyused Texas asphalt-binders, namely, PG 64-22 and PG 76-22,from 20 different suppliers. Note that the selected asphalt-

2 Advances in Civil Engineering

binders in this study presented the same low-temperaturegrade of −22, i.e., PG XX-22.

2.2. Additives and Plant-Mix Extracted Asphalt-Binders.With asphalt-binders extracted from plant-mix materials aswas the case in this study, HMA mix additivities such asrecycled asphalt pavement (RAP) and recycled asphaltshingles (RAS) usually tend to stiffen/harden the combinedasphalt-binder by increasing the proportion of the agedasphalt-binder in the total asphalt-binder blend [29–31].,is often results in increasing the stiffness of the asphalt-binder blend, which could potentially reduce the ductility(i.e., low ER values) of the asphalt-binder. ,us, in additionto the potential to reduce ductility and ultimately the HMAcracking resistance, these additives also have the potential toimpact the consistency and magnitude of the ER parameterof the plant-mix extracted asphalt-binders including thelaboratory test ER data variability. Note that the use of RAP/RAS additives in HMA mixes has become a commonpractice due partly to their economic and environmentalbenefits [29–31]. However, detailed evaluation of the RAP/RAS effects including chemical and volumetric analysis wasoutside the scope of this paper as the study’s focus was on thematerial-source effects.

3. Study Matrix Plan

,e study plan is comprised of using the Texas flexiblepavements and overlays database, namely, the Texas DataStorage System (DSS), as the primary data source. ,e DSS,the ductility test, asphalt-binders, and the statistical methodsused to analyze the data are discussed subsequently.

3.1. Data Source (the Texas DSS). Maintained in the readilyaccessible Microsoft Access® platform, the Texas DSS wascommissioned in 2010 to serve as an ongoing long-termdatabase for Texas flexible pavements and overlays[26, 32–34]. At the time of writing this paper, the DSS iscomprised of 115 in-service highway test sections withcomprehensive laboratory and field performance data thatincludes design, construction, layer material properties(both laboratory and field measured), traffic, climate,existing distresses for overlays, and field performance. ,eDSS’s extensive material properties include the laboratorymeasured asphalt-binder ER data from the Ductilometertest, which is the subject of this paper [26, 32].

In addition to the processed and analyzed data (in MS®Access format), the DSS has an accompanying raw datastorage system (namely, the Texas RDSS) that contains allthe corresponding raw data/files. ,ese raw data, in theRDSS, can be reprocessed and reanalyzed as needed. Fulldetails of the Texas DSS and RDSS can be found in Walubitaet al.’s work [26, 32–35].

3.2.�e Ductility Test: Elastic Recovery (ER). As per the DSStest plan, the ductility test was performed using a Ductil-ometer device according to the ASTM 6084 [22]

specification on plant-mix extracted asphalt-binders. Acentrifugal extraction method with a chlorinated solvent wasused for extracting the asphalt-binders from the preheatedloose HMA (plant-mix) which were hauled directly from thefield construction sites and treated as rolling thin film oven(RTFO) residues [26, 32–34]. ,ree specimens per asphalt-binder type/grade per source were conditioned in a bath at10°C (50°F) for about 1 hour prior to testing. After 1-hour10°C water-bath conditioning, the ductility test was thenconducted at a specimen elongation rate of 5 cm/min until a20 cm fixed elongation was obtained and held in this po-sition for 5min. ,ereafter, the specimens were cut at themidpoint into two halves and left undisturbed in the 10°Cwater bath for about 1 hour to allow recovery. ,e ductilitytest configuration and asphalt-binder specimens before andafter testing, as conducted during the DSS study, are shownin Figure 1 [26, 28, 33].

After 1 hour, both halves of the asphalt-binder specimenwere carefully (manually) adjusted to touch each other toallow for measurement of the total specimen length. ,epercentage elastic recovery (ER) was then determined usingthe following equation [22, 26, 28, 32–34]:

ER �e − x

e % . (1)

In equation (1), ER is the recovered elasticity (%), erepresents the original elongation of the specimen (cm), andx is defined as the elongation of the specimen (cm), at thecompletion of the specified recovery time (≈1 hour), with thesevered ends just touching each other [26, 28, 32–34].

3.3. Materials and Asphalt-Binders. As extracted from theDSS [26], 20 asphalt-binder sources/suppliers covering twocommonly used Texas PG XX-22 asphalt-binders (namely,PG 64-22 and PG 76-22) were statistically evaluated. ,easphalt-binders, with the suppliers donated as “Source01thru to Source20” for impartial anonymity, are listed inTable 1.

Note that all the asphalt-binders in Table 1 have the samelow-temperature grade, namely, −22, i.e., PG XX-22. Fur-thermore, 90% of the corresponding HMA mixes arecomprised of RAP and/or RAS additives.,erefore, a similarreference datum was assumed for the asphalt-binder sour-ces. However, as previously mentioned, detailed evaluationof the RAP/RAS effects including their age and chemical/volumetric analysis was not in the scope of this paper as thestudy’s focus was on the material-source effects.

3.4. Statistical Methods Used. For evaluating the data con-sistency, variability, and differences among the differentasphalt-binder sources/suppliers, the following statisticalmethods were employed:

(i) Standard MS® Excel descriptive statistics such asaverage (Avg) and CoV for assessing the data con-sistency, variability, and quality

(ii) ANOVA and Tukey’s HSD analysis for assessing thedifferences among the sources/suppliers

Advances in Civil Engineering 3

A CoV threshold of 30% (i.e., CoV≤ 30%) was used inthis study as a measure of test data consistency and vari-ability with the following subdesignations as suggested in theliterature: (a) CoV≤ 10% (excellent), (b) 10%

Although within the 30% CoV threshold [26], Source15(PG 64-22) exhibited more test data variability, with a CoVof 27.26%, that is, rated as marginal variability. By contrast,Table 2 shows that Source18 (PG 76-22) exhibited the besttest data consistency, with the smallest CoV value of 0.01%.Given that all the asphalt-binders in Table 1 have the samelow-temperature grade (i.e., −22) with most of them havingRAP/RAS additives, the results in Table 2 suggest thatmaterial source has a significant effect on the asphalt-binderER properties and data variability.

4.2. Performance Ranking. ,e ductility (ER) test is pri-marily used to evaluate and quantify the recoverability ofasphalt-binders after elastic elongation [19]. ,eoretically,higher values of the ER (ductility) in asphalt-bindersquantitatively represent better cracking resistance potential[10, 28]. With this consideration and based on the ER su-periority ranking in Table 2, the best source in terms ofpotential for cracking resistance is Source18, followed bySource19, both of which are PG 76-22 asphalt-binders.Source04, Source08, and Source09 present the worst per-formance with the lowest ER values, all of which are PG 64-22 asphalt-binders.

4.3. Data Quality and Consistency. ,e ER results in Table 2represent an average of three replicates per source per as-phalt-binder type/grade and, thus, permitted the statisticalassessment of data variability through CoV analysis, with30% used as the threshold, i.e., CoV≤ 30% [26, 38]. ,e ERresults exhibit CoV values lower than 30%, which represents

acceptable repeatability and data consistency, partly at-tributed to good workmanship, proper machine calibration,the use of trained operators, simplicity of the test, etc. [26].In general, the lower the CoV value, the better the consis-tency (i.e., lower variability) and data quality. Sources as-sociated with the lowest CoV values are Source18, Source02,and Source05, which represent the best sources in terms oftest data consistency and possibly better asphalt-binderquality. In fact, Source18 with PG 76-22, a typically polymer-modified asphalt-binder, ranks top (1st) in terms of con-sistency and test data quality, i.e., lowest CoV value. On theother hand, Source15, Source03, and Source13 present thehighest CoV values but lower than 30% and, thus, acceptabletest data consistency and reliability.

According to the ASTM D6084 specification [22], astandard deviation (Stdev) of 0.91% (i.e., Stdev≤ 0.91%) for a“single-operator precision” and an acceptable range of twotest results of 2.6% are recommended for the ER parameterfor PG 64-22 asphalt-binders. For PG 76-22, which are oftenpolymer-modified, the thresholds used were Stdev≤ 0.56and 1.60% for the acceptability, respectively [22]. Table 3shows the range of the maximum (Max) and minimum(Min) as well as the corresponding Stdev values.

Table 3 shows that eight of the 17 PG 64-22 sources (47%of the PG 64-22 sources) met the ASTM 2.60% acceptablerange with Source05 having the lowest value of 0.97%. Onthe other hand, only six of the 17 sources (i.e., 35%) met theASTM Stdev requirement of 0.91%, with a minimum Stdevof 0.49% for Source05. For the PG 76-22 asphalt-bindersources, only Source18met the ASTM thresholds, suggestingthe need to improve material consistency and quality in

Table 2: ER test results.

RTFO residue Test temperature� 10°C, elongation rate� 5 cm/minSource Asphalt-binder ER value (%) ER ranking CoV value (%) CoV rankingSource01 PG 64-22 23.50 16 2.13 5Source02 PG 64-22 36.67 8 1.57 2Source03 PG 64-22 27.10 12 23.78 19Source04 PG 64-22 17.67 20 14.24 16Source05 PG 64-22 26.79 13 1.83 3Source06 PG 64-22 27.84 11 6.27 12Source07 PG 64-22 25.40 14 2.38 7Source08 PG 64-22 23.01 18 2.81 9Source09 PG 64-22 21.17 19 17.32 17Source10 PG 64-22 29.65 10 7.20 13Source11 PG 64-22 32.04 9 7.80 14Source12 PG 64-22 59.00 3 3.39 10Source13 PG 64-22 41.00 7 18.41 18Source14 PG 64-22 23.49 17 3.40 11Source15 PG 64-22 24.32 15 27.26 20Source16 PG 64-22 51.00 4 1.96 4Source17 PG 64-22 50.00 5 2.58 8Source18 PG 76-22 72.00 1 0.01 1Source19 PG 76-22 65.19 2 2.25 6Source20 PG 76-22 46.14 6 10.42 15CoV: coefficient of variation; ER: elastic recovery; PG: performance-graded; RTFO: rolling thin film oven.

Advances in Civil Engineering 5

Source19 and Source20. ,e highest ER range (15.00%) andStdev (7.55%) recorded are for Source13 (PG 64-22), whichultimately does not meet the ASTM specification [22].

5. Statistical Analysis and Material-Source Effects

Statistical analyses were performed to ascertain if the sourceswere statistically significantly different (or not) based on theER parameter. Boxplots, ANOVA analysis, and Tukey’s HSDpairwise comparisons are presented in this section of thepaper.

5.1. Boxplots and ER Data. A boxplot is a standardizedmethod of graphically displaying and distinguishing datadistribution and detecting outlier presence, for any discretedata set [41]. From Figure 2, it is clearly seen that the truemedians differ for all the sources, with ER values between17% and 72% (on the vertical Y-axis). Source03, Source13,Source15, and Source20 present the greatest variability withwide boxes. Source01, Source02, Source05, and Source18, onthe other hand, present the lowest variability with tightlycompressed boxes. ,ese observations confirm the CoVresults in Table 2. Also, except for Source12, Source18,Source17, Source07, and Source08, most of the sourcespresent an asymmetrical (skewed) distribution of data withlarge whiskers. Statistically, large whiskers infer to large datadispersion and, consequently, high data variability.

In general, the boxplots in Figure 2 display two statisticalrepresentations and interpretations of the data [41, 42]. If theboxes overlap in the vertical orientation, it means that the

sources are statistically indifferent and vice versa. For in-stance, Source16 and Source17 are indifferent but are sta-tistically different from Source01 and Source18. Narrowboxes and shorter whiskers infer to high data consistencyand low variability, and vice versa [41, 42]. ,us, Source01exhibits better data consistency than Source13. Similarly,while Source01 and Source18 are both associated with highdata consistency and quality, they are statistically signifi-cantly different in terms of the ER magnitude.

5.2. ANOVA and Tukey’s HSD. ANOVA was performedusing an open-source statistical software R [43] at 95% CL(i.e., α� 5.0%� 0.05 for 95% CL) in terms of the p values.Interpretively, if p value is less than α, that is, p value < 0.05,then there might be some potential statistical differencesamong the sources/suppliers with respect to that particularparameter and vice versa [39]. Similarly, if the F value isgreater than the critical F, then there is a significant statisticaldifference among the sources. ,e results of ANOVAanalysis are summarized in Table 4.

From Table 4, the statistical F value of 72.0668 is sig-nificantly higher than the critical F of 1.8529, hence con-firming that there are statistically significant differencesamong the sources [39]. Table 4 also shows a probabilityvalue (p value) of 5.139x10− 25, which is considerably lowerthan the 0.05 significance level, meaning that, at 95% CL,there is at least one source among the 20 sources that couldbe statistically different from the others.

Although the ANOVA analysis provides a first insightinto the statistical differences on a whole population amongthe asphalt-binder sources, it cannot provide exactly where

Table 3: ER range and standard deviation.

RTFO residue Test temperature� 10°C, elongation rate� 5 cm/minSource Asphalt-binder Min ER Max ER Range Standard deviation (stdev)Source01 PG 64-22 23.00% 24.00% 1.00% 0.50%Source02 PG 64-22 36.00% 37.00% 1.00% 0.58%Source03 PG 64-22 20.00% 32.58% 12.58% 6.44%Source04 PG 64-22 15.00% 20.00% 5.00% 2.52%Source05 PG 64-22 26.36% 27.33% 0.97% 0.49%Source06 PG 64-22 26.04% 29.53% 3.49% 1.75%Source07 PG 64-22 24.80% 26.01% 1.21% 0.61%Source08 PG 64-22 22.34% 23.63% 1.29% 0.65%Source09 PG 64-22 19.05% 25.40% 6.35% 3.67%Source10 PG 64-22 27.19% 31.02% 3.83% 2.14%Source11 PG 64-22 29.15% 33.48% 4.33% 2.50%Source12 PG 64-22 57.00% 61.00% 4.00% 2.00%Source13 PG 64-22 34.00% 49.00% 15.00% 7.55%Source14 PG 64-22 22.91% 24.40% 1.49% 0.80%Source15 PG 64-22 19.00% 31.75% 12.75% 6.63%Source16 PG 64-22 50.00% 52.00% 2.00% 1.00%Source17 PG 64-22 48.71% 51.29% 2.58% 1.29%Source18 PG 76-22 72.00% 72.00% 0.00% 0.00%Source19 PG 76-22 63.50% 66.04% 2.54% 1.47%Source20 PG 76-22 40.64% 49.53% 8.89% 4.81%

,reshold [22]� — — PG 64-22≤ 2.60%PG 76-22≤1.60%PG 64-22≤ 0.91%PG 76-22≤ 0.56%

PG: performance-graded; ER: elastic recovery; CoV: coefficient of variation; Max: maximum;Min: minimum; numbers in red text: did not meet ASTMD6084specification [17].

6 Advances in Civil Engineering

the differences are. Hence, a Tukey Post Hoc Test (HSD) wasalso conducted to check which specific sources are different.,is test is a statistical tool used to determine if the rela-tionship between two sets of data is statistically significant bybuilding confidence intervals with an α level of significancefor all possible pairwise comparisons based on the followinghypotheses [27]:

H0 : μi � μj, versusH1 : μi ≠ μj, (2)

where H0 is the null hypothesis and H1 is the alternativehypothesis. A “True-False” methodology was proposed todenote that the differences in the ER results between eachpair of the asphalt-binder source were high enough to beconsidered statistically different at a 95% CL. ,ese True-False results are listed in Table 5, where “True” means thatthe paired sources are statistically significantly different andvice versa for “False,” that is, similar.

For the 190 possible source-pairs, 50% (95 source-pairs)were statistically different (i.e., “True” response in Table 5).Source18 presents the major statistical differences fromother sources, with 18 “True” responses and just one “False”response, followed by Source12, Source17, and Source19,with statistical differences between 16 and 17 sources (i.e., 16and 17 “True” responses). On the other hand, Source03 andSource 05 show the least “True” responses, with statistical

differences for seven sources (i.e., seven “True” responses)and 12 “False” responses, which means a statistical indif-ference with 12 sources.

Overall, Table 5 illustrates the variability of the ER testdata among the different sources. Ultimately, this indicatesthe sensitivity nature of the ER parameter to material-sourceeffects for plant-mix extracted asphalt-binders, that is, RTFOresidues.

6. Synthesis and Discussion of the Results

Figure 3 shows the graphical spreads associated with theasphalt-binder sources in terms of the ER parameter anddata variability (CoV).,e figure comprises a mean value foreach source and the overall average (Avg) incorporating allthe 20 asphalt-binder sources.

From Figure 3(a), the ER values range from about 17% to75%, with only three sources (i.e., Source12, Source18, andSource19) meeting the ER ≥ 59.00% threshold for crackingresistance potential of the corresponding HMAmix [10, 28].,e overall average ER is 36.15%. Twelve out of 20 sourcesfall below the average value. In terms of the asphalt-bindercomparisons, Source18 and Source19, which are all PG 76-22 asphalt-binders, exhibited the highest values, as theo-retically expected. ,e lowest ER value at 17.67% wasrecorded for PG 64-22 from Source04. Apparently, all theHMA mixes associated with PG 64-22 asphalt-binders werecomprised of RAP and RAS additives; see Table 1.,erefore,it is possible that these additives contributed to the low ERvalues of some PG 64-22 asphalt-binders such as Source04,which may not have been the case for Source12, Source16,and Source17 that exhibit ER values around 50% and rankedin the 3th, 4rd, and 5th positions in terms of performancesuperiority (Table 2).

Among the 20 asphalt-binder sources evaluated in thisstudy, only Source08 (PG 64-22), Source18 (PG 76-22), andSource20 (PG 76-22) had no RAP additive. However,Source08 had 3% RAS, while Source18 and Source20 bothhad 1% of lime. Based on the results presented in Sections 4and 5 of this paper, it was verified that asphalt-bindersclassified as PG 76-22 provided the overall highest ER values,ranging between 46.14% and 72%, respectively. In the case ofSource20, it should be highlighted that it presented thelowest ER value among the PG 76-22 asphalt-binders testedand also yielded the highest CoV (10.42%) value, that is, hadthe highest variability. ,is could have been probably due tothe presence of cellulose fibers that may have chemicallyinteracted with the chlorinated solvent during the extractionprocess, thus impacting the ER results, in terms of both theER magnitude and data variability [44, 45].

In analyzing the PG 64-22 asphalt-binders, an average of31.74% ER value was obtained, which represents almost halfthe average ER value (61.11%) of the PG 76-22 asphalt-binders. ,is may be explained from a rheological point ofview. PG 64-22 is two high-temperature grades below PG 76-22 in terms of rutting performance but at the same low-temperature grade of −22. ,is means that both asphalt-binders are theoretically expected to have the same responsebehavior in terms of controlling low-temperature cracking at

S1 S10

S11

S12

S13

S14

S15

S16

S17

S18

S19 S2 S20 S3 S4 S5 S6 S7 S8 S9

20

30

40

50

60

70

Source

ER (%

)

ER% boxplot @95%CL

Figure 2: Boxplot analysis.

Table 4: ANOVA results.

ANOVA results @ 95% CL |α� 5.0%� 0.05

Df Sum sq Mean sq F value Pr(>F)(p value) Critical F

Source 19 1.4397 0.07577 72.0668 5.139E-25 1.8529Residuals 40 0.0421 0.00105Total 59 1.4818Total 59 1.4818CL: confidence level; Df: degrees of freedom; Mean sq: squares mean; Pr: pvalue (probability); Sum sq: summation of squares; F: Fisher test; F val-ue> critical F: statistical difference exists and vice versa; Pr� p-value < α:statistical difference exists and vice versa.

Advances in Civil Engineering 7

Table 5: Tukey HSD results (True-False) @ 95%CL.

S-P T-F S-P T-F S-P T-F S-P T-F S-P T-FS10≠S1 False S13≠S11 False S2≠S13 False S19≠S16 True S8≠S19 TrueS11≠S1 False S14≠S11 False S20≠S13 False S2≠S16 True S9≠S19 TrueS12≠S1 True S15≠S11 False S3≠S13 True S20≠S16 False S20≠S2 FalseS1≠S1 True S16≠S11 True S4≠S13 True S3≠S16 True S3≠S2 FalseS14≠S1 False S17≠S11 True S5≠S13 True S4≠S16 True S4≠S2 TrueS15≠S1 False S18≠S11 True S6≠S13 True S5≠S16 True S5≠S2 FalseS16≠S1 True S19≠S11 True S7≠S13 True S6≠S16 True S6≠S2 FalseS17≠S1 True S2≠S11 False S8≠S13 True S7≠S16 True S7≠S2 TrueS18≠S1 True S20≠S11 True S9≠S13 True S8≠S16 True S8≠S2 TrueS19≠S1 True S3≠S11 False S15≠S14 False S9≠S16 True S9≠S2 TrueS2≠S1 True S4≠S11 True S16≠S14 True S18≠S17 True S3≠S20 TrueS20≠S1 True S5≠S11 False S17≠S14 True S19≠S17 True S4≠S20 TrueS3≠S1 False S6≠S11 False S18≠S14 True S2≠S17 True S5≠S20 TrueS4≠S1 False S7≠S11 False S19≠S14 True S20≠S17 False S6≠S20 TrueS5≠S1 False S8≠S11 False S2≠S14 True S3≠S17 True S7≠S20 TrueS6≠S1 False S9≠S11 True S20≠S14 True S4≠S17 True S8≠S20 TrueS7≠S1 False S13≠S12 True S3≠S14 False S5≠S17 True S9≠S20 TrueS8≠S1 False S14≠S12 True S4≠S14 False S6≠S17 True S4≠S3 FalseS9≠S1 False S15≠S12 True S5≠S14 False S7≠S17 True S5≠S3 FalseS11≠S10 False S16≠S12 False S6≠S14 False S8≠S17 True S6≠S3 FalseS12≠S10 True S17≠S12 False S7≠S14 False S9≠S17 True S7≠S3 FalseS13≠S10 True S18≠S12 True S8≠S14 False S19≠S18 False S8≠S3 FalseS14≠S10 False S19≠S12 False S9≠S14 False S2≠S18 True S9≠S3 FalseS15≠S10 False S2≠S12 True S16≠S15 True S20≠S18 True S5≠S4 FalseS16≠S10 True S20≠S12 True S17≠S15 True S3≠S18 True S6≠S4 TrueS17≠S10 True S3≠S12 True S18≠S15 True S4≠S18 True S7≠S4 FalseS18≠S10 True S4≠S12 True S19≠S15 True S5≠S18 True S8≠S4 FalseS19≠S10 True S5≠S12 True S2≠S15 True S6≠S18 True S9≠S4 FalseS2≠S10 False S6≠S12 True S20≠S15 True S7≠S18 True S6≠S5 FalseS20≠S10 True S7≠S12 True S3≠S15 False S8≠S18 True S7≠S5 FalseS3≠S10 False S8≠S12 True S4≠S15 False S9≠S18 True S8≠S5 FalseS4≠S10 True S9≠S12 True S5≠S15 False S2≠S19 True S9≠S5 FalseS5≠S10 False S14≠S13 True S6≠S15 False S20≠S19 True S7≠S6 FalseS6≠S10 False S15≠S13 True S7≠S15 False S3≠S19 True S8≠S6 FalseS7≠S10 False S16≠S13 False S8≠S15 False S4≠S19 True S9≠S6 FalseS8≠S10 False S17≠S13 False S9≠S15 False S5≠S19 True S8≠S7 FalseS9≠S10 False S18≠S13 True S17≠S16 False S6≠S19 True S9≠S7 FalseS12≠S11 True S19≠S13 True S18≠S16 True S7≠S19 True S9≠S8 FalseS-P: source-pair; T-F: True-False; S4≠S1: Source04 is not similar to Source01; True: the paired sources are statistically significantly different; False: the pairedsources are statistically indifferent (i.e., similar).

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

ER (%

)

ER (%)Avg ER = 36.15%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20No. of data points (sources)

ER ≥ 59.00%

(a)

0.00

5.00

10.00

15.00

20.00

25.00

30.00

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

COV

(%)

No. of data points (sources)

CoV (%)Avg CoV = 7.85%

CoV ≤ 30%

(b)

Figure 3: ER-CoV graphical plots.

8 Advances in Civil Engineering

−22°C no matter the high-temperature asphalt-binder grade.However, their ER response behavior, as evident in Table 2and Figure 3, was different in terms of their quantitativemagnitudes.,is could be partially explained by the fact thatPG grades higher than 70°C typically incorporate some typeof modifiers, that is, modified. In fact, PG asphalt-bindersthat differ in the high- and low-temperature specification by90°C or more generally require some sort of modification[46].

On the other hand, among the 17 PG 64-22 asphalt-binder sources evaluated, only three sources (namely,Source12, Source16, and Source17) had ER values ex-ceeding 50%. Similarly, three other sources (namely,Source02, Source11, and Source13) had their ER valuesranging between 30% and 50%, respectively. ,e possiblereasons for these variations include the following as-sumptions: (i) in spite of 16 out of 17 PG 64-22 sourcescontaining RAP and/or RAS with an almost similarcontent, the RAP and RAS came from different sources;(ii) the asphalt-binders from RAP/RAS mixes may havehad different levels of aging and residual asphalt-binderconcentration; and (iii) depending on the asphalt-bindersource (chemical composition) and RAP/RAS type, theremight have been different degrees of blending during theHMA production process, aspects which could haveimpacted the ER data variability. To verify these as-sumptions, further physical and chemical testing in-cluding SARA (i.e., Saturate, Aromatic, Resin, andAsphaltene) fractional analyses and RAP/RAS evaluationsare recommended [47, 48]. However, as previouslymentioned, detailed evaluation of the RAP/RAS effectsnor the SARA fractional analysis was not in the scope ofthis paper.

In terms of the ER data variability, Figure 3(b) showed arange of CoV values between 0.00% and 30.00%, indicatingan excellent to marginal data consistency with none ex-ceeding the 30% threshold [26, 33]. An average CoV value of7.85% was obtained and 14 out of 20 sources present ex-cellent data consistency below this average [47]. Foursources present good data consistency (i.e., 10%

quality through, among others, enhancing theirproduction and quality-control processes for thesesources.

Overall, the study highlighted the sensitivity nature ofthe ER parameter with respect to evaluating the effects ofmaterial sources and suppliers on the plant-mix extractedasphalt-binders’ ER properties based on Ductilometermeasurements. ,e test results and findings confirmed thatmaterial source has an impact on the ER property of asphalt-binders. From the study findings, it can also be concludedthat one has to be cautious of the material-source effect onthe rheological properties, grading, and, ultimately, theperformance of the asphalt-binders. ,e study also suggeststhat, in as much as performance superiority (and costs ofcourse) is a very crucial issue in deciding the asphalt-bindersource and supplier, consistency and quality aspects cannotbe ignored. ,at is, in addition to cost considerations,material-source effect should be holistically viewed andassessed from both performance (rheological properties)and quality (consistence) perspectives.

8. Research Significance and Future Studies

,is study has yielded technically informative data related toasphalt-binders extracted from plant-mix materials (i.e.,RTFO residue) as a function of material-source effects. Morestudies of this nature are recommended to further supple-ment and substantiate the results and findings reportedherein. ,ese types of studies are fundamental in trying tounderstand the field performance of asphalt-binders and theresultant HMA mixes considering the material-source ef-fects. Quite often asphalt-binders of the same type/grademay behave differently, displaying markedly different agingcharacteristics, water or stripping susceptibilities, fatigueresistance, low-temperature strength, flexibility, etc. ,at is,although asphalt-binders from different sources and sup-pliers may register the same type/grade classification, itshould not always be assumed that they would automaticallyexhibit exactly the same response behavior and performance.Depending on the source and supplier, some performancedifferences may exist even though it is the same asphalt-binder type/grade classification. ,us, studies of this natureare imperative to distinctively and quantitatively charac-terize the ductility and ER response behavior of asphalt-binders from different sources and suppliers, even thoughthey may be having the same low-temperature grade.

Overall, the study contributes to enriching the literatureon the material-source effects on asphalt-binders’ ERproperties, consistency, variability, and data quality. Inparticular, the study highlights the sensitivity nature of theasphalt-binder ER parameter to material-source effects.However, the population size was limited to 20 differentTexas asphalt-binder sources and suppliers for PG 64-22 andPG 76-22 asphalt-binders. ,erefore, future follow-upstudies should cover more sources/suppliers and asphalt-binder types/grades. In addition, the study approach couldinclude a detailed evaluation of the RAP/RAS effects in-cluding chemical and volumetric analysis which were not in

the scope of this paper. Laboratory tests covering physical,rheological, and chemical tests such as viscosity, penetration,ring and ball, fatigue performance, the multiple stress creep-recovery (MSCR), and SARA (Saturate, Aromatic, Resin,and Asphaltene) fractions could also be conducted tosupplement these findings. Other aspects for future studiescould also include a comparative documentation of theproduction processes and quality-control practices engagedby at the different asphalt-binder sources and suppliersevaluated in this study.

Data Availability

All data used to support the findings of this study areavailable from the corresponding author upon request.

Disclosure

,e contents of this paper, which do not constitute astandard, reflect the views of the authors who are solelyresponsible for the facts and accuracy of the data presentedherein and do not necessarily reflect the official views orpolicies of any agency or institute.

Conflicts of Interest

,e authors declare that they have no conflicts of interest.

Acknowledgments

,e authors acknowledge the Texas Department of Trans-portation (TxDOT) and the Federal Highway Administra-tion (FHWA) for supporting the Texas DSS project thatvaluably served as the data source for the work presented inthis paper. Special thanks also go to Brett Haggerty (PE) forpioneering the DSS work (Project No: 0-6658) and for histechnical guidance.

References

[1] K. Lill, K. Kontson, A. N. Khan, P. Pan, and S. A. M. Hesp,Comparison of Physical and Oxidative Aging Tendencies forCanadian and Northern European Asphalt Binders,S. Goodman, Ed., in Proceedings of the 64th CanadianTechnical Asphalt Association Annual Conference, pp. 369–385, Canadian Technical Asphalt Association, Montreal, QC,Canada, 2019.

[2] X. Hu and L. F. Walubita, “Influence of asphalt-binder sourceon CAM mix rutting and cracking performance: a laboratorycase study,” International Journal of Pavement Research andTechnology, vol. 8, no. 6, p. 419, 2015.

[3] A. E. Alvarez, L. V. Espinosa, S. Caro, E. J. Rueda, J. P. Aguiar,and L. G. Loria, “Differences in asphalt binder variabilityquantified through traditional and advanced laboratorytesting,” Construction and Building Materials, vol. 176, 2018.

[4] C. J. Robinette, T. M. Breakah, R. C. Williams, andJ. P. Bausano, “Evaluation of the variability of |E∗| with fieldprocured hot mix asphalt concrete mixtures,” Road Materialsand Pavement Design, vol. 11, 2010.

[5] A. Sreeram and Z. Leng, “Variability of rap binder mobi-lisation in hot mix asphalt mixtures,” Construction andBuilding Materials, vol. 201, 2019.

10 Advances in Civil Engineering

[6] J. J. Adams, M. D. Elwardany, J. J. Planche, R. B. Boysen, andJ. F. Rovani, “Diagnostic techniques for various asphalt re-fining and modification methods,” Energy & Fuels, vol. 33,no. 4, pp. 2680–2698, 2019.

[7] J. P. Planche, M. D. Elwardany, J. J. Adams, R. Boysen, andJ. Rovani, Linking Binder Characteristics with Performance:�e Recipe to Cope with Changes in Bitumen Binder Quality,in Proceedings of the XXVIth Permanent International Asso-ciation of Road Congresses (PIARC), World Road Congress,Abu Dhabi, UAE, 2019.

[8] G. G. Al-Khateeb and N. M. Al-Akhras, “Properties ofPortland cement-modified asphalt binder using Superpavetests,” Construction and Building Materials, vol. 25, 2011.

[9] European Asphalt Pavement Association, ,e Asphalt PavingIndustry, A Global Perspective, NAPA (USA), EAPA, Brus-sels, Belgium, 2011.

[10] J. Zhang, A. N. M. Faruk, P. Karki, I. Holleran, X. Hu, andL. F. Walubita, “Relating asphalt binder elastic recoveryproperties to HMA cracking and fracture properties,” Con-struction and Building Materials, vol. 121, 2016.

[11] G. S. L. F. Walubita, L. Fuentes, L. Sang-Ick, O. Guerrero,E. Mahmoud, and B. Naik, “Correlations and preliminaryvalidation of the laboratory monotonic overlay test (OT) datato reflective cracking performance of in-service field highwaysections,” Construction and Building Materials, vol. 121029,2020.

[12] S. Hu, F. Zhou, and T. Scullion, “Factors that affect crackingperformance in hot-mix asphalt mix design,” TransportationResearch Record: Journal of the Transportation ResearchBoard, vol. 2210, no. 1, pp. 37–46, 2011.

[13] Y. Tan, L. Shan, Y. Richard Kim, and B. S. Underwood,“Healing characteristics of asphalt binder,” Construction andBuilding Materials, vol. 27, no. 1, pp. 570–577, 2012.

[14] ASTM, “D113-17, standard method of test for ductility ofasphalt materials,” ASTM International, vol. 11, pp. 7–10,2017.

[15] P. S. Kandhal, “Low-temperature ductility in relation topavement performance,” in Low-Temperature Prop. Bitum.Mater. Compact. Bitum. Paving Mix., C. R. Marek, Ed.,pp. 95–106, ASTM International, West Conshohocken, PA,USA, 1977.

[16] Y. Ruan, R. R. Davison, and C. J. Glover, “An investigation ofasphalt durability: relationships between ductility and rheo-logical properties for unmodified asphalts,” Petroleum Scienceand Technology, vol. 21, 2003.

[17] ASTM International, ASTM D946/D946M-15 StandardSpecification for Penetration-Graded Asphalt Binder for Use inPavement Construction, West Conshohocken, PA, USA, 2015.

[18] I. Isailović, M. P. Wistuba, and A. Cannone Falchetto, “In-vestigation on mixture recovery properties in fatigue tests,”Road Materials and Pavement Design, vol. 19, pp. 1230–1240,2018.

[19] F. Ma, X. Luo, Z. Huang, and J. Wang, “Characterization ofrecovery in asphalt binders,” Materials, vol. 13, no. 4, p. 920,2020.

[20] A. Nawaz Khan, M. Akentuna, P. Pan, and S. A. M. Hesp,“Repeatability, reproducibility, and sensitivity assessments ofthermal and fatigue cracking acceptance criteria for asphaltcement,” Construction and Building Materials, vol. 243, Ar-ticle ID 117956, 2020.

[21] H. Tabatabaee, C. Clopotel, A. Arshadi, andH. Bahia, “Criticalproblems with using the asphalt ductility test as a performanceindex for modified binders,” Transportation Research RecordJournal of the Transportation, vol. 2370, no. 1, pp. 84–91, 2013.

[22] ASTM International,ASTMD6084/D6084M-18 Standard TestMethod for Elastic Recovery of Asphalt Materials by Ductil-ometer, West Conshohocken, PA, USA, 2018.

[23] C. S. Clopotel and H. U. Bahia, “Importance of elastic re-covery in the DSR for binders and mastics,” EngineeringJournal, vol. 16, no. 4, pp. 99–106, 2012.

[24] A. Shenoy, “A dynamic oscillatory test that fulfills the ob-jective of the elastic recovery test for asphalt binders,” Ma-terials and Structures, vol. 41, no. 6, pp. 1039–1049, 2008.

[25] J. J. Adams, M. D. Elwardany, J. P. Planche, R. B. Boysen, andJ. F. Rovani, “Diagnostic techniques for various asphalt re-fining and modification methods,” Energy and Fuels, vol. 16,2019.

[26] L. F. Walubita, S. I. Lee, A. N. M. Faruk, T. Scullion,S. Nazarian, and I. Abdallah, Texas Flexible Pavements andOverlays: Year 5 Report—Complete Data Documentation,Austin, TX, USA, 2017, http://tti.tamu.edu/documents/0-6658-3.pdf.

[27] J. Montañez, S. Caro, D. Carrizosa, A. Calvo, and X. Sánchez,“Variability of the mechanical properties of Reclaimed As-phalt Pavement (RAP) obtained from different sources,”Construction and Building Materials, vol. 230, 2020.

[28] J. Zhang, G. S. Simate, S. I. Lee, S. Hu, and L. F. Walubita,“Relating asphalt binder elastic recovery properties to HMAcrack modeling and fatigue life prediction,” Construction andBuilding Materials, vol. 111, 2016.

[29] N. Tran, A. Taylor, and R. Willis, “Effect of rejuvenator onperformance properties of HMA mixtures with high RAPcontents,” Asphalt Pavements, vol. 2, pp. 1661–1670, 2014.

[30] W. Mogawer, A. Booshehrian, S. Vahidi, and A. Austerman,“Evaluating the effect of rejuvenators on the degree ofblending and performance of high RAP, RAS, RAP/RASmixtures,” RoadMaterials and Pavement Design, vol. 14, 2013.

[31] A. Bonicelli, P. Calvi, G. Martinez-Arguelles, L. Fuentes, andF. Giustozzi, “Experimental study on the use of rejuvenatorsand plastomeric polymers for improving durability of highRAP content asphalt mixtures,” Construction and BuildingMaterials, vol. 155, pp. 37–44, 2017.

[32] L. F. Walubita, G. Das, E. Espinoza et al., Texas FlexiblePavements and Overlays: Data Analysis Plans and ReportingFormat, Austin, TX, USA, 2012, http://tti.tamu.edu/documents/0-6658-P3.pdf.

[33] L. F. Walubita, G. Das, E. Espinoza et al., Texas FlexiblePavements and Overlays: Year 1 Report-Test Sections, DataCollection, Analyses, and Data Storage System, Austin, TX,USA, 2012, http://tti.tamu.edu/documents/0-6658-1.pdf.

[34] L. F. Walubita, R. Hassan, S. I. Lee et al., Data Collection andPopulation of the Database (�e DSS and RDSSP), https://static.tti.tamu.edu/tti.tamu.edu/documents/0-6658-P5.pdf, 2014.

[35] L. F. Walubita, S. I. Lee, A. N. M. Faruk et al., Texas FlexiblePavements and Overlays: Calibration Plans for M-E Modelsand Related Software, https://static.tti.tamu.edu/tti.tamu.edu/documents/0-6658-P4.pdf, 2013.

[36] G. F. Reed, F. Lynn, and B. D. Meade, “Use of coefficient ofvariation in assessing variability of quantitative assays,”Clinical and Vaccine Immunology, vol. 9, no. 6, pp. 1235–1239,2002.

[37] Q. Brook, Lean Six Sigma & Minitab : �e Complete ToolboxGuide for All Lean Six Sigma Practitioners, OPEX ResourcesLtd, Winchester, UK, 2010.

[38] P. Kriz, K. Sokol, D. Meskas, and S. Maria, “Statistical ap-proach to dsr-pav test improvement,” in Proceedings of theInternational Society for Asphalt Pavements 2016 Symposium,Jackson Hole, WY, USA, July 2016.

Advances in Civil Engineering 11

http://tti.tamu.edu/documents/0-6658-3.pdfhttp://tti.tamu.edu/documents/0-6658-3.pdfhttp://tti.tamu.edu/documents/0-6658-P3.pdfhttp://tti.tamu.edu/documents/0-6658-P3.pdfhttp://tti.tamu.edu/documents/0-6658-1.pdfhttps://static.tti.tamu.edu/tti.tamu.edu/documents/0-6658-P5.pdfhttps://static.tti.tamu.edu/tti.tamu.edu/documents/0-6658-P5.pdfhttps://static.tti.tamu.edu/tti.tamu.edu/documents/0-6658-P4.pdfhttps://static.tti.tamu.edu/tti.tamu.edu/documents/0-6658-P4.pdf

[39] S. M. Ross, Introduction to Probability and Statistics for En-gineers and Scientists, 2004.

[40] C. E. Brown, “Coefficient of variation,” Applied MultivariateStatistics in Geohydrology and Related Science, Springer,Berlin, Germany, 1998.

[41] M. B. Adam, B. I. Babura, and K. Gopal, “Range-box plottingrelating to discrete distribution,” MATEMATIKA, vol. 34,no. 2, pp. 187–204, 2018.

[42] Understanding Boxplots-towards Data Science. https://towardsdatascience.com/understanding-boxplots-5e2df7bcbd51.

[43] R. P. ,e, “,e R project for statistical computing,” 2020,https://www.r-project.org/.

[44] T. Bai, Z.-a. Hu, X. Hu, Y. Liu, L. Fuentes, and L. F. Walubita,“Rejuvenation of short-term aged asphalt-binder using wasteengine oil,” Canadian Journal of Civil Engineering, vol. 47,no. 7, pp. 822–832, 2020.

[45] X. Hu, S. Fan, X. Li, P. Pan, L. Fuentes, and L. F. Walubita,“Exploring the feasibility of using reclaimed paper-basedasphalt felt waste as a modifier in asphalt-binders,” Con-struction and Building Materials, vol. 234, Article ID 117379,2020.

[46] Superpave Performance Grading-Pavement Interactive,https://pavementinteractive.org/reference-desk/materials/asphalt/superpave-performance-grading/.

[47] L. F. Walubita, “Comparison of fatigue analysis approachesfor predicting fatigue lives of hot-mix asphalt concrete(HMAC) mixtures,” Doctoral Dissertation, Texas A&MUniversity, Texakana, TX, USA, 2006.

[48] S. Ashoori, M. Sharifi, M. Masoumi, and M. MohammadSalehi, “,e relationship between SARA fractions and crudeoil stability,” Egyptian Journal of Petroleum, vol. 26, no. 1,pp. 209–213, 2017.

12 Advances in Civil Engineering

https://towardsdatascience.com/understanding-boxplots-5e2df7bcbd51https://towardsdatascience.com/understanding-boxplots-5e2df7bcbd51https://towardsdatascience.com/understanding-boxplots-5e2df7bcbd51https://www.r-project.org/https://pavementinteractive.org/reference-desk/materials/asphalt/superpave-performance-grading/https://pavementinteractive.org/reference-desk/materials/asphalt/superpave-performance-grading/